Research map / Detection and segmentation / Open-set and few-shot detection
Benchmarking Open-ended Segmentation
Cristina González, Santiago Rodriguez, Kevis-kokitsi Maninis, Jordi Pont-Tuset, Pablo Arbelaez
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use OPAL (contrastive region-text alignment training) instead of Osprey-7B | Osprey-7B · ADE20K val open-ended panoptic | PQ: 46.6 -> 48.8 | Improved |
| use OPAL instead of Osprey-7B | Osprey-7B · Cityscapes val open-ended panoptic | PQ: 50.2 -> 52.8 | Improved |
| use OPAL instead of Osprey-7B | Osprey-7B · Cityscapes val open-ended instance | AP: 31.3 -> 31.9 | Improved |
| contrastive loss in training | OPAL · ADE20K and Cityscapes panoptic robustness across seeds | output variance: reduces output variance by nearly 50% | Improved |
| lexical mapping with Gemini instead of GPT-4 (ablation) | OPAL · ADE20K val | PQ: 47.9 -> 48.0 | No clear effect |
| lexical mapping with Gemini instead of GPT-4 (ablation) | OPAL · Cityscapes val | PQ: 52.4 -> 48.3 | Hurt |
| lexical mapping with Gemini instead of GPT-4 (ablation) | Osprey · Cityscapes val | PQ: 49.0 -> 46.2 | Hurt |
| lexical mapping instead of Sentence-BERT mapping | evaluation protocol · human-verified Cityscapes disagreement subset | accuracy vs human label: 4.8 % -> 84.4 % | Improved |
| increase noun list size from 20% to full (ablation) | OPAL · ADE20K open-ended panoptic | lexical alignment curve: 6% gap between the full noun list and the 20% subset | Improved |
| MLLM-generated tags as vocabulary for MasQCLIP | MasQCLIP · ADE20K val open-ended panoptic | PQ: 38.6 -> 42.9 | Improved |
Closest papers by meaning
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.89)
- USE: Universal Segment Embeddings for Open-Vocabulary Image Segmentation (CVPR 2024 · similarity 0.88)
- OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts (NeurIPS 2025 · similarity 0.88)
- Open-Vocabulary Segmentation with Semantic-Assisted Calibration (CVPR 2024 · similarity 0.88)
- Training-Free Open-Vocabulary Segmentation with Offline Diffusion-Augmented Prototype Generation (CVPR 2024 · similarity 0.88)
- Training-free Open-Vocabulary Semantic Segmentation via Diverse Prototype Construction and Sub-region Matching (AAAI 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
